[Edu-sig] PyGeo distribution progress.

[Arthur]

I'm going to take the shot that some might consider this  on topic to
edu-sig.

Anyone not interested in the real life story of new (and happy) distutils
user, and my PyGeo distribution configuration progress, have no need to read
further.

Jason writes -

>YES please switch to distutils
>For the longterm health of PyGeo

>It's one of the most important changes to Python to make it friendly for
both
>newbies and pros. And that's no mean feat..

>Everyone will thankyou.

Guido writes -

>No, distutils is *not* overkill.  Please consider using distutils; it
>will make others' lives so much easier!  If you get stuck, I'll try to
>help.


Where this has led -

Considerable rethinking not only of the form of distribution of PyGeo, but
its organization, etc.

Where I am at.

For ceating any PyGeo script:

from pygeo import *

does it all.  A *considerable* improvement from where things were a week
ago.  Thanks Kirby for the help.

As to disutils:

got it down, and have a distro of standalone PyGeo that puts everything
nicely where it belongs by:

"python setup.py install"

The nice thing about distutils is that the setup.py script that is used to
"package" the distro is the exact script that is used by the enduser to
install.

But there's more.

>From this setup script I can create a self-installing Windows executable.

Who knew?

Distutils implicitly recognizing that the Windows and Linux communities
represent, in many ways, have  different users and therefore different
needs.  And it accomodates both well, easing the developers life. Good
stuff.

Thanks again, Guido, for getting me started down this road.

But this is still Art speaking, so of course there is a problem. And with
some "political" content.

The VPython distribution on which PyGeo depends chooses to be off in its own
world. In my view, to an extreme.

Among the bizarre things it does is overwrite the *Python's* root
documentation file, "index.html". Which means, among other strange effects,
that even after you might uninstall VPython, you will forever get VPython's
root help file, rather than Python's. *Which no longer exists*. Obviosuly
absurd, IMO.  And ill-mannered. IMO. And believe me I've raised this with
Mr. Sherwood - to no avail.

But even solving that, there are other issues.

Not only does VPython install in non-standard places, it includes Numeric,
but installs Numeric differently than Numeric installs itself on a
standalone basis.

So my plan is to go the step further.

Probaly just ask my users to install the "official" Numeric. There is a
Windows self installing executable. And I just tested the from source
shebang under Windows, the setup.py install, that calls my VC++ and compiles
and then installs in site-packages, and it worked quickly and flawlessly,
and I ended up with something I would consider "standard" placement of files
under the Python root.

And then include my own VPython installation with PyGeo.

Don't know  a lot about these licensing things, but if VPython can offer a
Numeric distribution that "mangles" its distribution, I can only assume that
I am within my rights to do something similar as to VPython.

The best choice, as far as I am confirmed, is to have VPython comply to with
some reasonable standards.  Apparently not in the cards.

ALL FWIW, I guess.

Art